400 Days of Fear & Greed: The Index Spent 72% of Them Frightened

We pulled every daily reading of the Crypto Fear & Greed Index from 1 July 2025 to 4 August 2026 — 400 consecutive observations.¹ The index sat on the fear side of its own scale on 290 of those days. That is 72.5%. It printed “Extreme Greed” exactly once. Its highest reading all year was 79, on a scale that runs to 100.

The distribution

• Extreme Fear (0–24) — 171 days — 42.8%
• Fear (25–44) — 119 days — 29.8%
• Neutral (45–55) — 51 days — 12.8%
• Greed (56–75) — 58 days — 14.5%
• Extreme Greed (76–100) — 1 day — 0.3%

The mean reading was 32.3 and the median 26. A quarter of the year came in at 17 or below; a tenth at 11 or below. The lowest print was 5. The index also ran 94 consecutive days on the fear side without a single break, beginning 19 January 2026. Split the window in half and the first ninety days averaged 59.4 while the last ninety averaged 24.3.

What an index is supposed to do

A sentiment gauge earns its keep by telling you where today sits relative to normal. To do that it has to have a normal — a level it returns to, with excursions either side. This one does not behave that way. Over 400 days it used the bottom half of its range four times as often as the top, and left the top fifth of the scale entirely untouched. On the evidence of this year, the readable statement is not “the market is frightened.” It is “the price has been falling,” which you could have established by looking at the price.

Our view is blunt: an oscillator that does not oscillate is a trend indicator wearing a dial. That is not a fatal criticism — trend indicators are useful and this one is free, legible and updated daily. But it is routinely presented as a contrarian tool, on the theory that extremes mark turning points, and a series that spent 42.8% of the year at an “extreme” cannot support that reading. If nearly half of all days are extreme, the word has stopped doing any work.

There is also the composition problem. The index is a weighted blend of volatility, momentum, volume, social media activity, dominance and search trends. Several of those inputs are functions of price by construction. Momentum is price. Volatility is price. Dominance is price divided by other prices. Building a sentiment index largely out of price and then presenting it as an independent read on sentiment is circular, and the shape of this year’s distribution is what that circularity looks like from the outside.

What this does not show

One window, and a directional one. Bitcoin fell 44.1% over roughly the same period, so an index that tracks price was always going to spend the year in the basement; a rising year would produce the mirror image, and neither on its own proves the index is uninformative. The honest test needs several full cycles, and 400 days is not that.

We have not verified the index’s construction beyond its publisher’s own description of the inputs and their weights, and we cannot reproduce it from raw data. The category boundaries — 24, 44, 55, 75 — are the publisher’s, and the count of “extreme” days would change if they were drawn elsewhere. Nothing here says the index predicts anything, in either direction; that is a separate measurement and we have run it separately.

We will re-pull the full series on 3 November 2026 and report whether the top half of the scale gets used.